Adaptive model-based seasonal adjustment of time series: An Empirical Comparison with XII-ARIMA and SIGEX

Several model-based approaches have been proposed in recent years for adjusting and decomposing time series data. Using real world data, this paper presents results of a large scale empirical comparison of the XII-ARIMA and SIGEX procedures to DESAEP, a new adaptive model-based method that combines stochastic and deterministic effects. The comparison reveals no substantial inconsistencies in seasonally adjusted values produced by the 3 methods. As for the magnitude in revisions in both concurrent and forecasted seasonally adjusted values, an overall reduction by a factor of 2 to 3 was obtained with DESAEP depending on the level of variability in the data.